Transferable Deep Reinforcement Learning for Cross-Domain Navigation: from Farmland to the Moon
Shreya Santra, Thomas Robbins, Kazuya Yoshida
TL;DR
This work tackles cross-domain autonomous navigation by training a DRL policy in a farmland-like environment and evaluating it zero-shot in a lunar-like setting. Using Proximal Policy Optimization (PPO) within an NVIDIA Isaac Sim-based pipeline, the authors deploy a four-wheeled Adam rover with 12-dimensional observations and two-DOF actions to learn goal-directed navigation and obstacle avoidance. The study demonstrates that a terrestrial-trained policy can achieve substantial transfer to extraterrestrial terrains, recording an average lunar success rate of 46.69% (best case 73.33%) without any lunar retraining, underscoring the potential for reducing retraining costs in space missions. While promising, the results also reveal sensitivity to obstacle placement and environment realism, motivating further improvements in simulation fidelity, architecture, and eventual sim-to-real validation for reliable planetary exploration autonomy.
Abstract
Autonomous navigation in unstructured environments is essential for field and planetary robotics, where robots must efficiently reach goals while avoiding obstacles under uncertain conditions. Conventional algorithmic approaches often require extensive environment-specific tuning, limiting scalability to new domains. Deep Reinforcement Learning (DRL) provides a data-driven alternative, allowing robots to acquire navigation strategies through direct interactions with their environment. This work investigates the feasibility of DRL policy generalization across visually and topographically distinct simulated domains, where policies are trained in terrestrial settings and validated in a zero-shot manner in extraterrestrial environments. A 3D simulation of an agricultural rover is developed and trained using Proximal Policy Optimization (PPO) to achieve goal-directed navigation and obstacle avoidance in farmland settings. The learned policy is then evaluated in a lunar-like simulated environment to assess transfer performance. The results indicate that policies trained under terrestrial conditions retain a high level of effectiveness, achieving close to 50\% success in lunar simulations without the need for additional training and fine-tuning. This underscores the potential of cross-domain DRL-based policy transfer as a promising approach to developing adaptable and efficient autonomous navigation for future planetary exploration missions, with the added benefit of minimizing retraining costs.